Self-supervised speaker embeddings

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26230%2F19%3APU134182" target="_blank" >RIV/00216305:26230/19:PU134182 - isvavai.cz</a>

Výsledek na webu

<a href="https://www.isca-speech.org/archive/Interspeech_2019/pdfs/2842.pdf" target="_blank" >https://www.isca-speech.org/archive/Interspeech_2019/pdfs/2842.pdf</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.21437/Interspeech.2019-2842" target="_blank" >10.21437/Interspeech.2019-2842</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Self-supervised speaker embeddings

Popis výsledku v původním jazyce

Contrary to i-vectors, speaker embeddings such as x-vectors are incapable of leveraging unlabelled utterances, due to the classification loss over training speakers. In this paper, we explore an alternative training strategy to enable the use of unlabelled utterances in training. We propose to train speaker embedding extractors via reconstructing the frames of a target speech segment, given the inferred embedding of another speech segment of the same utterance. We do this by attaching to the standard speaker embedding extractor a decoder network, which we feed not merely with the speaker embedding, but also with the estimated phone sequence of the target frame sequence. The reconstruction loss can be used either as a single objective, or be combined with the standard speaker classification loss. In the latter case, it acts as a regularizer, encouraging generalizability to speakers unseen during training. In all cases, the proposed architectures are trained from scratch and in an endto- end fashion. We demonstrate the benefits from the proposed approach on the VoxCeleb and Speakers in the Wild Databases, and we report notable improvements over the baseline.

Název v anglickém jazyce

Self-supervised speaker embeddings

Popis výsledku anglicky

Contrary to i-vectors, speaker embeddings such as x-vectors are incapable of leveraging unlabelled utterances, due to the classification loss over training speakers. In this paper, we explore an alternative training strategy to enable the use of unlabelled utterances in training. We propose to train speaker embedding extractors via reconstructing the frames of a target speech segment, given the inferred embedding of another speech segment of the same utterance. We do this by attaching to the standard speaker embedding extractor a decoder network, which we feed not merely with the speaker embedding, but also with the estimated phone sequence of the target frame sequence. The reconstruction loss can be used either as a single objective, or be combined with the standard speaker classification loss. In the latter case, it acts as a regularizer, encouraging generalizability to speakers unseen during training. In all cases, the proposed architectures are trained from scratch and in an endto- end fashion. We demonstrate the benefits from the proposed approach on the VoxCeleb and Speakers in the Wild Databases, and we report notable improvements over the baseline.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2019

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Proceedings of Interspeech

ISBN

—

ISSN

1990-9772

e-ISSN

—

Počet stran výsledku

5

Strana od-do

2863-2867

Název nakladatele

International Speech Communication Association

Místo vydání

Graz

Místo konání akce

INTERSPEECH 2019

Datum konání akce

15. 9. 2019

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000831796403001